Astrobe Forum

Thanks for your help. I'm impressed by the small size and power of Oberon. Is it practical to use it for machine learning?

What quantity of data would you be processing?

What rate of processing would you require (e.g. bytes / second)?

Is the data being supplied in realtime or from a storage system?

How would the results be applied?

What algorithms would you be using to process the data?

The more information you can give about the application you want to implement, the more accurate is the answer I can give you (and vice versa).

My intention is to develop an image recognition system for plant identification. The goal is to control and remove unwanted plants from agricultural land, pastures and eventually other lands. Initially the system would be trained with static images. Then it would be used to monitor fields for plant species using static images. If the system is fast enough it can use real time images from drones. If the confidence level is high enough the system would be used for spot spraying or mechanical removal of unwanted plants.
There are limited and expensive systems now that can do some of this. I would like a more generalized system that would have a core function of plant identification and add on modules for specific functions. Such a system could be mass produced.

This is not the sort of system I would contemplate building from scratch. Instead I recommend that you investigate an established Machine Learning framework like Apple's Core ML:

https://developer.apple.com/documentation/coreml

frankl wrote:Thanks for your help. I'm impressed by the small size and power of Oberon. Is it practical to use it for machine learning?

I have experience with Oberon (Component Pascal) for machine learning. I made simple multilayer network for classification of kinetic curves.
https://molpit.org/page/56
Image processing is more complicated issue, however in principle there should not be the big difference.